Hello,
I have been trying to run Myrna on a 5 node hadoop cluster with CentOS following the directions at:
Myrna was installed on the master node on NFS, so all the slave nodes can also access the application. I set all the environment variables in my .bash_profile file and checked to make sure they are available on all nodes. SSH is passwordless. Here is the command that I used:
$MYRNA_HOME/myrna_hadoop \
--preprocess \
--input=hdfs://myrna_pa/example/yeast/small.manifest \
--output=hdfs://myrna_pa/example/yeast/output_small \
--reference=hdfs://myrna_refs_pa/yeast_ensembl_67.jar \
--streaming=/usr/lib/hadoop-mapreduce/hadoop-streaming.jar
But I get the following error:
Myrna expects 'bowtie' to be at path /home/thad/myrna_pa/myrna-1.2.0/bin/linux64/bowtie on the workers
Myrna expects 'Rscript' to be at path /home/thad/myrna_pa/myrna-1.2.0/R/bin/Rscript on the workers
Myrna expects 'fastq-dump' to be at path /home/thad/myrna_pa/sratoolkit.2.1.16-centos_linux64/bin/fastq-dump on the workers
Could not parse Hadoop version: "mapreduce/hadoop"
Strangely, when I did not the set up right such as env vars in .bash_profile etc, I was getting this same error. Nothing seems to fix it. Any help will be appreciated.
Other potentially useful information is that I have installed Cloudera Distribution for Hadoop (CDH), version 4. One reason for the problem might be that Myrna does not work with this version. The last error message (red) suggests that the version string was not parsable by Myrna code (MyrnaIface.pm) so it died.
Thanks,
- Pankaj
I have been trying to run Myrna on a 5 node hadoop cluster with CentOS following the directions at:
Myrna was installed on the master node on NFS, so all the slave nodes can also access the application. I set all the environment variables in my .bash_profile file and checked to make sure they are available on all nodes. SSH is passwordless. Here is the command that I used:
$MYRNA_HOME/myrna_hadoop \
--preprocess \
--input=hdfs://myrna_pa/example/yeast/small.manifest \
--output=hdfs://myrna_pa/example/yeast/output_small \
--reference=hdfs://myrna_refs_pa/yeast_ensembl_67.jar \
--streaming=/usr/lib/hadoop-mapreduce/hadoop-streaming.jar
But I get the following error:
Myrna expects 'bowtie' to be at path /home/thad/myrna_pa/myrna-1.2.0/bin/linux64/bowtie on the workers
Myrna expects 'Rscript' to be at path /home/thad/myrna_pa/myrna-1.2.0/R/bin/Rscript on the workers
Myrna expects 'fastq-dump' to be at path /home/thad/myrna_pa/sratoolkit.2.1.16-centos_linux64/bin/fastq-dump on the workers
Could not parse Hadoop version: "mapreduce/hadoop"
Strangely, when I did not the set up right such as env vars in .bash_profile etc, I was getting this same error. Nothing seems to fix it. Any help will be appreciated.
Other potentially useful information is that I have installed Cloudera Distribution for Hadoop (CDH), version 4. One reason for the problem might be that Myrna does not work with this version. The last error message (red) suggests that the version string was not parsable by Myrna code (MyrnaIface.pm) so it died.
Thanks,
- Pankaj
Comment